Fast Colorization of Grayscale Images by Convolutional Neural Network
Swathy Titus, Jency N.M. Rena · 2018
Image colorization is the process of adding chrominance values to an input grayscale image. A grayscale image is the one which contains only the intensity values. Colorization is a very important but challenging task since a natural looking colour image has to be obtained from any grayscale input. Among the various colorization techniques, convolutional neural network based colorization is selected because of its ability to deal with image datasets. Unlike the previous techniques, neural network based colorization method is a fully automatic technique which does not require any human assistance. It exploits machine learning concept in which the computers find the solution by learning the datasets and arriving at a conclusion. It includes pre-processing, segmentation, feature extraction and model generation. The experimental results demonstrated that this method outperforms all the existing techniques. Image colorization is applicable in many areas such as colorization of old black and white photos, old movies and scientific images.